This paper explores nonlinear methods, inspired by the fractal theory for the analysis of the structure of music signals at multiple time scales, which is of importance both for their modeling and for their automatic computer based recognition. The m